Dynamic sensitive data desensitization method in cloud environment and related equipment

Abstract

The invention discloses a dynamic sensitive data desensitization method in a cloud environment and related equipment, and relates to the field of data desensitization, the method comprises the following steps: generating a task chain permission dependency graph according to role permission combination, the task chain permission dependency graph comprising permission inheritance and transmission relationships among task nodes, the role permission combination comprises a viewing permission level, a data operation permission and a timeliness constraint; according to the current task node position of the target user, predicting the next task operation of the target user in the task chain permission dependency graph, and pre-loading the target role permission corresponding to the next task operation; determining a preset dynamic desensitization rule of the target sensitive data according to the target role permission; and desensitizing the target sensitive data according to a preset dynamic desensitization rule in the data access process of the target sensitive data. By implementing the method, the processing efficiency of data desensitization can be improved.

Description

Technical Field

[0001] The present application relates to the field of data desensitization, and in particular to a method and related equipment for dynamic desensitization of sensitive data in a cloud environment. Background Art

[0002] With the rapid development of cloud computing technology, more and more companies are choosing to migrate their business systems to the cloud environment. In the cloud environment, the security protection of sensitive data of enterprises faces greater challenges, especially in multi-role collaborative work scenarios, where the access rights of different roles to sensitive data require fine-grained management and control.

[0003] In related technologies, sensitive data protection solutions usually use static data desensitization, that is, sensitive fields are fixedly masked or replaced when data is stored. At the same time, user access rights to sensitive data are managed through a pre-set role permission configuration table, and users need to perform identity authentication and permission verification when accessing data.

[0004] However, when users need to handle cross-system related business, each time they access a new business module, they need to re-authenticate their permissions and match data desensitization rules. This mechanism will cause the system to respond slowly and affect business processing efficiency. Summary of the invention

[0005] The present application provides a method and related equipment for dynamic desensitization of sensitive data in a cloud environment, which are used to improve the processing efficiency of data desensitization.

[0006] In the first aspect, the present application provides a method for dynamic desensitization of sensitive data in a cloud environment, which is applied to a dynamic desensitization system for sensitive data in a cloud environment. The method includes: generating a task chain permission dependency graph based on a role permission combination, wherein the task chain permission dependency graph includes the permission inheritance and transfer relationship between each task node, and the role permission combination includes viewing permission level, data operation permission and timeliness constraint; predicting the next task operation of the target user in the task chain permission dependency graph based on the current task node position of the target user, and preloading the target role permission corresponding to the next task operation; determining a preset dynamic desensitization rule for the target sensitive data based on the target role permission; and desensitizing the target sensitive data according to the preset dynamic desensitization rule during data access to the target sensitive data.

[0007] By adopting the above technical solution, a task chain permission dependency graph is generated according to the role permission combination, which includes the permission inheritance and transfer relationship between task nodes. The role permission combination includes viewing permission level, data operation permission and timeliness constraint. When the user performs a task operation, the next operation can be accurately predicted based on this graph and the corresponding permissions can be preloaded, thereby improving the system's response speed and the accuracy of permission management, ensuring that users can efficiently access and process sensitive data under the premise of legality and compliance, and enhancing the security and controllability of sensitive data in the cloud environment.

[0008] In combination with some embodiments of the first aspect, in some embodiments, before the step of generating a task chain permission dependency graph based on the role permission combination, the method also includes: obtaining the operation timing data of the target user, and extracting the standard work task chain of the target user based on the operation timing data; marking each task node of the standard work task chain with the role permission combination corresponding to the task node.

[0009] By adopting the above technical solution, before generating the task chain permission dependency graph, the target user's operation time series data is first obtained and the standard work task chain is extracted, and then the role permission combination is annotated for each task node. Through the analysis and processing of the operation time series data, the user's work mode and habits can be accurately captured, so that the generated standard work task chain is more in line with the actual work situation, and the role permission combination annotation of each task node further refines the permission management, so that the system can perform permission control and data desensitization more accurately according to the user's specific permission requirements at different task nodes.

[0010] In combination with some embodiments of the first aspect, in some embodiments, the step of obtaining the operation timing data of the target user and extracting the standard work task chain of the target user based on the operation timing data specifically includes: performing time series analysis on the operation timing data to identify repeated operation patterns in the operation timing data; determining an operation sequence in the repeated operation pattern whose occurrence frequency exceeds a preset threshold as a candidate task chain; performing task similarity calculation on the candidate task chains, merging the candidate task chains whose similarities are greater than a preset similarity threshold to obtain the standard work task chain of the target user, wherein the task similarity is determined by calculating the degree of matching of the operation type, operation object and operation time of each task node in the candidate task chain.

[0011] By adopting the above technical solution, when extracting the standard work task chain based on the operation time series data, the data is analyzed in time series to identify repeated operation patterns, and the operation sequences with a frequency exceeding the preset threshold are determined as candidate task chains, and then the standard work task chain is obtained by merging them by calculating the task similarity. This method makes full use of the historical data of user operations, and through scientific analysis methods, it mines representative workflows, so that the generated standard work task chain more objectively and accurately reflects the actual work situation of users.

[0012] In combination with some embodiments of the first aspect, in some embodiments, the step of generating a task chain permission dependency graph based on role permission combinations specifically includes: obtaining the role permission combination corresponding to each task node in the standard work task chain; analyzing the role permission combinations of adjacent task nodes in the standard work task chain to generate a permission correlation matrix; determining the permission inheritance relationship between task nodes based on the permission correlation matrix, and generating an initial permission dependency graph; analyzing the permission transfer relationship between non-adjacent task nodes based on the initial permission dependency graph, the permission transfer relationship indicating that permissions can be transferred through intermediate task nodes; updating the permission transfer relationship to the initial permission dependency graph to form a task chain permission dependency graph.

[0013] By adopting the above technical scheme, in the process of generating a task chain permission dependency graph according to the role permission combination, the role permission combination corresponding to each task node in the standard work task chain is first obtained, and then the role permission combination of adjacent task nodes is analyzed to generate a permission correlation matrix, and then the permission inheritance relationship between task nodes is determined and an initial permission dependency graph is generated, and then the permission transfer relationship between non-adjacent task nodes is analyzed and updated to the initial permission dependency graph, so that the task chain permission dependency graph can comprehensively and accurately reflect the complex permission relationship between task nodes, including direct permission inheritance and indirect permission transfer.

[0014] In combination with some embodiments of the first aspect, in some embodiments, the step of predicting the next task operation of the target user in the task chain permission dependency graph based on the current task node position of the target user specifically includes: obtaining the historical operation trajectory of the target user in the task chain permission dependency graph; analyzing the association pattern between the task nodes in the historical operation trajectory to obtain the association relationship between the task nodes; determining the predicted task operation set that the target user may perform based on the association relationship and the current task node position of the target user; and selecting the task node with the highest execution probability as the next task operation based on the historical execution frequency of each task node in the predicted task operation set.

[0015] By adopting the above technical scheme, the next task operation of the target user is predicted in the task chain permission dependency graph according to the current task node position of the target user. By obtaining the historical operation trajectory of the user in the task chain permission dependency graph, the association pattern between the task nodes is analyzed to obtain the association relationship, and then the predicted task operation set is determined in combination with the current task node position, and the task node with the highest execution probability is selected as the next task operation based on the historical execution frequency. The user's behavior can be predicted in advance and the corresponding preparations can be made for the system in advance, such as preloading the target role permissions.

[0016] In combination with some embodiments of the first aspect, in some embodiments, after the step of desensitizing the target sensitive data according to the preset dynamic desensitization rule during the data access process of the target sensitive data, the method also includes: obtaining historical audit log data of the target user within a preset time period, the historical audit log data including the target user's access time, access frequency, access location, data operation type and task node conversion record in the task chain permission dependency graph of the target sensitive data; training a machine learning model based on the historical audit log data to obtain a normal access behavior baseline model of the target user; evaluating the real-time access behavior of the target user according to the normal access behavior baseline model, and when the deviation between the real-time access behavior and the normal access behavior exceeds a preset threshold, adjusting the desensitization level in the preset dynamic desensitization rule to the preset level.

[0017] By adopting the above technical solution, after desensitizing the target sensitive data, the historical audit log data of the target user within the preset time period is obtained, and the machine learning model is trained based on this to obtain the normal access behavior baseline model, and then the real-time access behavior of the target user is evaluated according to the model. When the deviation between the real-time access behavior and the normal access behavior exceeds the preset threshold, the desensitization level in the preset dynamic desensitization rule is adjusted to the preset level, realizing real-time monitoring of user access behavior and dynamic adjustment of the desensitization level, and being able to timely detect abnormal access behavior and take corresponding protection measures.

[0018] In combination with some embodiments of the first aspect, in some embodiments, after the step of training a machine learning model based on the historical audit log data to obtain a normal data access behavior baseline model of the target user, the method also includes: collecting the current access data of the target user in real time, and inputting it into the normal data access behavior baseline model for calculation to obtain an abnormality degree score for the current access behavior; and adjusting the access rights to the target sensitive data according to the abnormality degree score.

[0019] By adopting the above technical solution, after training the machine learning model based on historical audit log data to obtain the normal data access behavior baseline model of the target user, the current access data of the target user is collected in real time and input into the model to calculate the abnormality degree score, and then the access rights to the target sensitive data are adjusted according to the abnormality degree score. When abnormal user behavior is detected, possible malicious operations or misoperations can be effectively prevented by reducing or restricting their access rights, thereby further strengthening the protection of sensitive data.

[0020] In the second aspect, an embodiment of the present application provides a dynamic desensitization system for sensitive data in a cloud environment, wherein the dynamic desensitization system for sensitive data in a cloud environment comprises: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code comprises computer instructions, and the one or more processors call the computer instructions to enable the dynamic desensitization system for sensitive data in the cloud environment to execute the method described in the first aspect and any possible implementation manner of the first aspect.

[0021] In the third aspect, an embodiment of the present application provides a computer program product comprising instructions. When the above-mentioned computer program product is run on a dynamic desensitization system for sensitive data in a cloud environment, the above-mentioned dynamic desensitization system for sensitive data in the cloud environment executes the method described in the first aspect and any possible implementation method of the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions. When the instructions are executed on a dynamic desensitization system for sensitive data in a cloud environment, the dynamic desensitization system for sensitive data in the cloud environment executes the method described in the first aspect and any possible implementation method of the first aspect.

[0023] It is understandable that the sensitive data dynamic desensitization system in the cloud environment provided by the second aspect, the computer program product provided by the third aspect, and the computer storage medium provided by the fourth aspect are all used to execute the method provided by the embodiment of the present application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method, which will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. This application generates a task chain permission dependency graph based on role permission combinations, which includes permission inheritance and transfer relationships between task nodes. The role permission combination includes viewing permission levels, data operation permissions, and time constraints. When users perform task operations, they can accurately predict the next operation based on this graph and preload corresponding permissions, thereby improving the system's response speed and the accuracy of permission management, ensuring that users can efficiently access and process sensitive data under the premise of legality and compliance, and enhancing the security and controllability of sensitive data in the cloud environment.

[0025] 2. This application extracts standard work task chains based on operation time series data, performs time series analysis on the data to identify repeated operation patterns, determines operation sequences with a frequency exceeding a preset threshold as candidate task chains, and then merges them by calculating task similarity to obtain standard work task chains. This method makes full use of the historical data of user operations, and uses scientific analysis methods to mine representative workflows, so that the generated standard work task chains more objectively and accurately reflect the actual work situation of users.

[0026] 3. This application predicts the next task operation of the target user in the task chain permission dependency graph according to the current task node position of the target user. By obtaining the historical operation trajectory of the user in the task chain permission dependency graph, the association pattern between the task nodes is analyzed to obtain the association relationship, and then the predicted task operation set is determined in combination with the current task node position. Based on the historical execution frequency, the task node with the highest execution probability is selected as the next task operation. This can predict the user's behavior in advance and make corresponding preparations for the system in advance, such as preloading the target role permissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flowchart of a method for dynamic desensitization of sensitive data in a cloud environment in an embodiment of the present application; Figure 2 It is another flowchart of the method for dynamic desensitization of sensitive data in a cloud environment in an embodiment of the present application; Figure 3 It is another flowchart of the method for dynamic desensitization of sensitive data in a cloud environment in an embodiment of the present application; Figure 4 It is a schematic diagram of the physical device structure of the sensitive data dynamic desensitization system in the cloud environment in an embodiment of the present application. DETAILED DESCRIPTION

[0028] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification of the present application, the singular expressions "one", "a kind of", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more of the listed items.

[0029] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0030] For ease of understanding, the following is a description of the process of the method provided by this implementation. Figure 1 , which is a flow chart of a method for dynamic desensitization of sensitive data in a cloud environment in an embodiment of the present application.

[0031] S101. Generate a task chain permission dependency graph based on the role permission combination. The task chain permission dependency graph includes the permission inheritance and transfer relationship between each task node. The role permission combination includes viewing permission level, data operation permission and timeliness constraint.

[0032] Among them, the role permission combination represents the set of permissions granted to the user in a specific task, including attributes such as data access level, operation type and validity period. The task chain permission dependency graph refers to a directed graph structure that describes the permission relationship between task nodes, and is used to represent the inheritance and transfer rules of permissions. Task nodes represent independent operation units in the workflow. The permission inheritance relationship refers to the subordinate relationship that the high-level task node contains the permissions of the low-level task node. The permission transfer relationship represents the mechanism by which permissions are indirectly transferred through intermediate nodes. The viewing permission level refers to the classification standard for the visible range of data. Data operation permissions represent the scope of authorization for operations such as adding, deleting, modifying and checking data. Time constraints are used to limit the effective time period of permissions.

[0033] During the system initialization phase, it is necessary to establish a task chain permission dependency graph as the basic framework for permission management. Specifically, the system first obtains the permission configuration information of all task nodes, including the permission level, operation permissions, and time constraints of each node. Then, the permission relationship between adjacent task nodes is analyzed to determine the inheritance direction and transfer rules of permissions. For permission inheritance, the system establishes a parent-child node relationship to ensure that the child node inherits the basic permissions of the parent node. For permission transfer, the system defines the transfer path and transfer conditions to achieve cross-node permission sharing. Finally, all permission relationships are integrated into a complete dependency graph structure.

[0034] In some embodiments, the generation of the task chain permission dependency graph can be achieved in a variety of ways. Optionally, first build a permission attribute table for the task node and define the basic permission configuration for each node; then analyze the business relationship between the nodes to determine the permission inheritance and transfer rules; finally, use the graph data structure to store all permission relationships. Optionally, a bottom-up approach is adopted to first define the most basic task node permissions; then build the permission hierarchy layer by layer through permission extension rules; finally, form a complete permission dependency network. It is understandable that other methods can also be used to generate the task chain permission dependency graph, such as based on historical data analysis, expert rule definition, etc., which are not limited here.

[0035] S102: predict the next task operation of the target user in the task chain permission dependency graph according to the current task node position of the target user, and preload the target role permission corresponding to the next task operation.

[0036] Among them, the target user refers to the system user who currently needs to perform permission prediction. The task node position refers to the workflow node that the user is currently executing. The next task operation refers to the subsequent task that the user may perform after the current node is completed. The target role permission refers to the permission configuration required for the predicted task node. Permission preloading means preparing the permission data required for the next operation in advance. The task chain permission dependency graph is a directed graph model that describes the permission relationship between task nodes. The prediction process is a calculation process that infers the user's next behavior based on historical data and current status.

[0037] When the user is performing a task, the system needs to predict the user's next operation in real time and prepare the required permissions in advance. Specifically, the system first obtains the information of the task node the user is currently in, including the node type, execution status, and context data. Then, based on the node relationship in the task chain permission dependency graph and the user's historical operation mode, the most likely next operation is predicted. For the predicted task node, the system loads the corresponding permission data from the permission configuration library in advance, including access rights, operation rights, and validity period restrictions.

[0038] In some embodiments, task operation prediction and permission preloading can be implemented in a variety of ways. Optionally, use machine learning methods to analyze user historical operation records and establish a behavior prediction model; select the most likely next operation based on the model prediction results; load the corresponding permission configuration into the memory cache in advance. Optionally, define task flow rules based on the rule engine; analyze the execution conditions and business parameters of the current node; determine the next operation based on rule matching and preload permissions. It is understandable that other methods can also be used to implement task operation prediction and permission preloading, such as statistical probability models, expert systems and other methods, which are not limited here.

[0039] S103. Determine preset dynamic desensitization rules for target sensitive data according to the target role authority.

[0040] Among them, the target role permission refers to the permission configuration granted to the user in a specific task node. Target sensitive data refers to business data that needs to be protected and desensitized. The preset dynamic desensitization rule refers to the data desensitization strategy automatically generated according to the permission level. Dynamic desensitization refers to the mechanism of adjusting the desensitization strategy in real time according to the access scenario and permission level. Desensitization rules are used to define the processing method of data fields. The rule determination process refers to the mapping process of converting permission requirements into specific desensitization strategies. Sensitive data fields represent data attributes that need to be protected. Field processing method refers to the specific method of deforming or hiding sensitive data.

[0041] When the system receives a data access request, it needs to determine the appropriate desensitization strategy based on the user's role permissions. Specifically, the system first parses the data access level, operation permissions, and time constraints in the target role permissions. Then selects a matching basic rule set from the desensitization rule template library. For each sensitive data field, the system selects a corresponding desensitization method based on the permission level, such as masking, replacement, encryption, etc. At the same time, consider the special requirements of the business scenario and adjust the desensitization parameters and exception rules. Finally, a complete dynamic desensitization rule configuration is generated, including field mapping, processing methods, and parameter settings.

[0042] In some embodiments, dynamic desensitization rules can be determined in a variety of ways: optionally, read permission configuration information through a rule engine to match predefined desensitization rule templates; adjust desensitization parameters according to permission levels and business needs; generate rule execution scripts and deploy them to the data access layer. Optionally, a configuration mapping method is used to establish a correspondence table between permission levels and desensitization policies; parse user permission attributes to select appropriate policy combinations; dynamically generate desensitization rule configurations and cache them in memory. It is understandable that other methods can also be used to determine dynamic desensitization rules, such as machine learning methods, expert system reasoning and other technologies, which are not limited here.

[0043] S104. During the data access process to the target sensitive data, the target sensitive data is desensitized according to the preset dynamic desensitization rule.

[0044] The data access process represents the complete process of the system reading and returning sensitive data. The preset dynamic desensitization rules refer to a predetermined set of data processing strategies. Desensitization processing is used to represent the operation of deforming or hiding sensitive data. The target sensitive data refers to the original data content that needs to be protected. The processing process represents the specific steps to execute the desensitization rules. Data deformation refers to the method of changing the display form of data through a specific algorithm. Data hiding refers to the technology of replacing sensitive information with placeholders. Rule execution refers to the process of applying the desensitization strategy to specific data.

[0045] When a user requests access to sensitive data, the system needs to perform real-time desensitization before returning the data. Specifically, the system first intercepts the query request at the data access layer and loads the desensitization rule configuration for the current session. Then it parses the query result set and identifies the data fields that need to be desensitized. For each sensitive field, the system calls the corresponding desensitization processor to perform data conversion operations. The processor selects the appropriate algorithm based on the rule configuration, such as regular replacement, encryption transformation, randomization, etc. Finally, the desensitized data is reassembled into a result set and returned to the user.

[0046] In some embodiments, the desensitization of sensitive data can be achieved in a variety of ways: optionally, query results are obtained through a data access interceptor, and the fields that need to be desensitized are parsed; the corresponding desensitization algorithm is called to perform data conversion; the processed data is repackaged and returned. Optionally, a parallel processing framework is used to slice the data set for concurrent processing; the corresponding desensitization rules are applied to each slice; the processing results are merged and consistency checked. It is understandable that other methods can also be used to achieve the desensitization of sensitive data, such as streaming processing, asynchronous processing and other technical solutions, which are not limited here.

[0047] The following is a more detailed description of the process of the method provided by this implementation. Figure 2 , which is another flow chart of the method for dynamic desensitization of sensitive data in a cloud environment in an embodiment of the present application.

[0048] S201: Acquire operation time series data of a target user, perform time series analysis on the operation time series data, and identify repeated operation patterns in the operation time series data.

[0049] Target users refer to specific user accounts that need to access sensitive data in the cloud environment. Operational time series data refers to the record of all operations performed by users in the system and the time when they occurred, including operations such as logging into the system, viewing files, editing data, submitting for approval, and the timestamp information of their specific occurrence. Time series analysis is a data analysis method that performs statistics and pattern recognition on data arranged in chronological order. Repeated operation patterns refer to fixed operation combinations that appear multiple times in the user's operation sequence.

[0050] The system first continuously records every operation of the user through the log collection module, and the record content includes information such as operation ID, operation type, operation object, and operation time. For the collected operation log data, the system uses a sliding window method to mine sequence patterns. The window size can be set to 8 hours (one working day) and the sliding step is 1 hour. In each window, a sequence pattern mining algorithm (such as the PrefixSpan algorithm) is used to identify frequently occurring operation sequences. During the identification process, the operation sequence is sorted by timestamp, and the operation combination with time sequence association is extracted by calculating the operation interval time, operation type conversion rules and other features.

[0051] S202: Determine an operation sequence in the repeated operation mode whose occurrence frequency exceeds a preset threshold as a candidate task chain.

[0052] A repetitive operation pattern refers to a fixed combination of operations that appear multiple times in time series data. The frequency of occurrence refers to the proportion of a specific operation sequence in the total operation records. The preset threshold is a frequency judgment standard predefined by the system to filter high-frequency operation sequences. A candidate task chain refers to a combination of operation sequences that may represent a user's standard workflow.

[0053] The system counts the frequency of the identified repetitive operation patterns, calculates the number of occurrences of each operation sequence in all operation records, and divides it by the total number of operation days to get the average daily frequency. The preset threshold can be set to 0.3, that is, if an operation sequence has an average daily occurrence frequency of more than 30%, it will be judged as a candidate task chain. During the statistical process, the system also considers the time span of the operation sequence, requiring that the operation sequence must maintain a high frequency for more than 3 consecutive months to ensure that a stable working mode is screened out rather than a temporary operation combination.

[0054] S203. Calculate task similarity for the candidate task chains, merge candidate task chains with similarity greater than a preset similarity threshold, and obtain a standard work task chain for the target user. The task similarity is determined by calculating the matching degree of the operation type, operation object, and operation time of each task node in the candidate task chain.

[0055] Task similarity is an indicator to measure the similarity between different candidate task chains. Operation type refers to the specific operation method on data, such as viewing, modifying, deleting, etc. Operation object refers to the specific data item or functional module being operated. Operation time refers to the time point when each operation occurs and its temporal relationship. The preset similarity threshold is the standard value for judging whether to merge task chains. The standard work task chain is the standardized workflow obtained after merging.

[0056] S204: Mark each task node of the standard work task chain with a role authority combination corresponding to the task node.

[0057] The standard work task chain refers to the standardized workflow sequence after identification and merging. Task nodes refer to each independent operation step in the workflow. The role permission combination includes three dimensions: viewing permission level (determines the visible scope of data), data operation permissions (such as reading, modifying, deleting, etc.) and timeliness constraints (the validity period of permissions). The annotation process is the process of establishing a mapping relationship between role permission information and task nodes. For example, for the task node "Customer Information Review", the corresponding role permission combination is {Viewing Level: Secondary Sensitive Data Visible, Operation Permission: Read-only, Timeliness: 9:00-18:00 on Working Days}.

[0058] The system implements permission labeling through a rule engine. First, a task-permission mapping rule base is established, which includes a basic permission rule set and a scenario permission rule set. The basic permission rule set defines the standard permission requirements for different types of task nodes. For example, the task of viewing customer information requires the permission to view basic customer information. The scenario permission rule set defines additional permission requirements for specific business scenarios, such as temporary authorization rules in emergency approval scenarios. For each task node, the system performs permission rule matching in turn to determine the minimum permission set required for the node. The permission labeling results are stored in the permission configuration table, which includes fields such as task node ID, permission combination ID, effective time, and expiration time.

[0059] S205: Obtain the role authority combination corresponding to each task node in the standard work task chain.

[0060] The role permission combination is a set of permission definitions, including permission type, permission level and constraint conditions. The permission type defines the operation mode on the data, such as reading, modifying, deleting, etc. The permission level determines the visible range of the data, such as fully visible, partially visible, etc. Constraints include time restrictions, geographical location restrictions, etc. The task node in the standard work task chain is the basic operation unit in the workflow, and each node requires a specific permission combination to execute. The system reads the permission information of the task node from the permission configuration database. Query the permission configuration table through the task node ID to obtain all permission records associated with the node. The permission record contains fields such as permission ID, permission type code, permission level code, and constraint conditions. The system parses the queried permission record into a permission object, which contains permission attributes and constraint rules. For each task node, the system builds a permission combination object, which contains all the permission information required for the node. The permission combination object is stored in a tree structure to facilitate subsequent permission inheritance and transfer analysis.

[0061] S206: Analyze the role authority combination of adjacent task nodes in the standard work task chain to generate a authority correlation matrix.

[0062] Adjacent task nodes refer to nodes that have a direct front-and-back relationship in the workflow. Role permission combination is the set of permissions required for each node. Permission correlation is an indicator to measure the closeness of the relationship between two permission combinations. The permission correlation matrix is ​​a two-dimensional matrix that records the degree of permission correlation between all nodes in the task chain. For example, the permission combinations of the two adjacent nodes "data entry" and "data review" have a high degree of correlation because the permissions of the review node include some permissions of the entry node.

[0063] The system uses the permission feature vector method to calculate the permission association. First, each permission combination is converted into a permission feature vector. The vector dimensions include data access level, operation type, time constraint and other features. For each pair of adjacent nodes in the task chain, the association score of its permission feature vector is calculated. The association calculation adopts the weighted summation method, and the weights of different permission features are: data access level 0.4, operation type 0.3, time constraint 0.3. Fill the calculated association into the N×N association matrix, where N is the number of nodes in the task chain. Each element aij in the matrix represents the permission association from node i to node j, and the value range is [0, 1].

[0064] S207. Determine the permission inheritance relationship between task nodes according to the permission association matrix, and generate an initial permission dependency graph.

[0065] The permission association matrix is ​​an N×N matrix that records the numerical values ​​of the permission association degree between task nodes. The permission inheritance relationship refers to the subordinate relationship that the permission of a task node contains the permission of another node. The initial permission dependency graph is a directed graph structure, where nodes represent tasks and edges represent permission inheritance relationships. For example, in the loan approval process, the "senior approval" node inherits all permissions of the "primary approval" node, forming an inheritance relationship.

[0066] The system constructs the initial permission dependency graph through the following steps: First, set the permission inheritance threshold to 0.8, traverse each element aij in the permission association matrix, and when aij is greater than the threshold, determine that there is a permission inheritance relationship between node i and node j. Use a graph data structure to store permission dependencies. Each graph node contains attributes such as task ID and permission combination. The graph edge represents the inheritance relationship, and the weight of the edge is the association value. The system uses an adjacency list to store the graph structure, which is convenient for subsequent graph traversal and analysis. Perform transitive closure calculations on the identified inheritance relationship to ensure the integrity of the inheritance relationship. Finally, the constructed graph structure is serialized and stored in the graph database.

[0067] S208. Analyze the permission transfer relationship between non-adjacent task nodes according to the initial permission dependency graph. The permission transfer relationship indicates that permissions can be transferred through intermediate task nodes.

[0068] The initial permission dependency graph is a directed graph that represents the direct permission inheritance relationship. Non-adjacent task nodes refer to nodes that are not directly connected in the task chain. The permission transfer relationship refers to the relationship chain in which permissions are indirectly transferred through intermediate nodes. Intermediate task nodes are transition nodes in the permission transfer process. For example, in the "data entry-data review-final review" process, the final review node can indirectly obtain some permissions of the data entry node through the data review node.

[0069] The system uses a graph traversal algorithm to analyze the authority transfer relationship. First, the Floyd-Warshall algorithm is used to calculate the shortest path between any two nodes, and the path length uses the inverse of the authority correlation as the edge weight. For each pair of non-adjacent nodes (i, j), if there is a path whose authority correlation product is greater than 0.6, it is considered that there is an authority transfer relationship between nodes i and j. The system records all intermediate nodes on the transfer path and calculates the actual authority range after the transfer. The authority transfer calculation uses the authority intersection method to ensure that the authority range will not expand during the transfer process. The analyzed transfer relationship is stored in the transfer relationship table, which contains information such as the source node, target node, intermediate node sequence, and transfer authority range.

[0070] S209. Update the permission transfer relationship to the initial permission dependency graph to form a task chain permission dependency graph. The task chain permission dependency graph includes the permission inheritance and transfer relationship between each task node. The role permission combination includes viewing permission level, data operation permission and timeliness constraint.

[0071] The permission transfer relationship is an indirect permission sharing mechanism implemented through intermediate nodes. The task chain permission dependency graph is a complete graph structure that includes all permission relationships of direct inheritance and indirect transfer. The viewing permission level defines the visible range of the data. The data operation permission specifies the specific operation type on the data. The timeliness constraint specifies the effective time range of the permission. For example, the permission combination of a node is {view level: level 3, operation permission: read and write, timeliness: valid within 24 hours}.

[0072] The system integrates the permission transfer relationship into the initial permission dependency graph. First, the records in the transfer relationship table are converted into edge relationships of the graph. The attributes of the edge include information such as the transfer type, transfer path, and the scope of permissions after transfer. For each transfer relationship, a dotted edge from the source node to the target node is added to the graph to represent an indirect permission relationship. The updated graph structure is stored using a property graph database. The node attributes include task information and permission combinations, and the edge attributes include relationship types (direct inheritance / indirect transfer) and permission transfer rules. The system attaches a permission calculator to each edge to dynamically calculate the actual available permission range. The final task chain permission dependency graph supports functions such as permission inheritance relationship query, permission transfer path analysis, and permission scope verification.

[0073] S210: Obtain the historical operation track of the target user in the task chain permission dependency graph.

[0074] The target user refers to a specific user account that needs to be predicted for permissions. The task chain permission dependency graph is a directed graph structure that describes the permission relationship between task nodes. The historical operation track is a record of the task sequence performed by the user in the system, including information such as task type, execution time, and execution order. For example, the historical operation track of a credit approval officer contains sequence data such as {time 1: application acceptance, time 2: document review, time 3: risk assessment, time 4: approval decision}.

[0075] The system extracts the target user's operation records for the last 90 days from the operation log database. The operation records contain fields such as operation ID, task node ID, operation timestamp, and operation status. The system first sorts the operation records by timestamp and constructs a time-series operation sequence. For each operation record, the task chain permission dependency graph is queried to obtain the complete information of the corresponding node, including node attributes and permission information. The system uses a sliding window method to group continuous operation records into operation sessions, each of which represents a complete business processing flow. Finally, all operation sessions are integrated into the user's historical operation trajectory and stored as an ordered sequence of task nodes.

[0076] S211, analyzing the association pattern between the task nodes in the historical operation track to obtain the association relationship between the task nodes.

[0077] The historical operation track is the time sequence of task nodes executed by the user. The association pattern refers to the regular combination of task nodes. The association relationship between task nodes describes the sequential dependency and conditional trigger relationship between nodes. For example, in the loan approval process, the "risk assessment" node is always executed after the "document review" node, and must be started within 4 hours after the "document review" is completed, which constitutes an association pattern.

[0078] The system uses a sequential pattern mining algorithm to analyze the operation trajectory. First, the SPADE (Sequential Pattern Discovery using Equivalence classes) algorithm is used to identify frequently occurring task sequence patterns, and the support threshold is set to 0.3. For each identified sequence pattern, the time interval distribution between task nodes is calculated, and the kernel density estimation method is used to obtain the optimal time window. The system constructs a task transition probability matrix, and the matrix elements represent the transition probability between nodes. The association rule mining algorithm is used to analyze the conditional dependency between nodes, and a rule set of the form "If task A is executed and condition C is met, the probability of executing task B within time window T is P" is obtained. Finally, the transition probability and conditional rules are combined to form a complete association relationship model.

[0079] S212: Determine a set of predicted task operations that the target user may perform based on the association relationship and the task node position where the target user is currently located.

[0080] The association relationship is a model that describes the transfer rules between task nodes. The current task node position refers to the task node that the user is currently executing. The predicted task operation set is a list of task nodes that the system predicts the user may execute next. For example, when the user completes the "loan material review" task, the predicted task operation set includes {risk assessment (probability 0.7), return supplementary materials (probability 0.2), direct rejection (probability 0.1)}.

[0081] The system predicts the next operation through the Markov decision process. First, obtain the status information of the user's current task node, including task completion, execution time, related business parameters, etc. Substitute the current status information into the association model to calculate the set of subsequent task nodes that meet the trigger conditions. For each candidate task node, the system calculates the transition probability score. The score calculation formula is: Final score = historical transition probability × 0.6 + time window matching × 0.2 + condition rule matching × 0.2. The system selects task nodes with a score greater than 0.4 to form a predicted task operation set. Each task node in the set contains a predicted probability value and trigger condition information. The prediction results are arranged in descending order of probability to facilitate subsequent permission preloading processing.

[0082] S213: Based on the historical execution frequency of each task node in the predicted task operation set, select the task node with the highest execution probability as the next task operation, and preload the target role authority corresponding to the next task operation.

[0083] The predicted task operation set is a list of possible subsequent task nodes predicted by the system. The historical execution frequency refers to the number of executions of the task node in the historical records. The execution probability is the probability of the task node being selected based on the historical data. The target role permissions are the permissions required for the task node to execute. Preloading refers to the process of preparing and caching permission data in advance. For example, in the credit approval process, if the historical execution frequency of the "risk assessment" node is 10 times per day, accounting for 70% of all subsequent operations, then it has the highest probability of being selected as the next task operation.

[0084] The system uses a weighted scoring method to select the optimal task node. For each node in the predicted task operation set, calculate the average daily execution frequency and execution time distribution in the last 30 days. Construct a scoring function: node score = historical execution frequency weight (0.5) × standardized frequency value + time consistency weight (0.3) × time matching score + business rule weight (0.2) × rule matching score. Select the task node with the highest score as the next step. Preload the permission configuration of the selected node, including: obtaining the complete permission definition from the permission configuration library, building a permission cache object, preheating the data access interface, and preparing the desensitization rule configuration. Store the preloaded permission data in the distributed cache system and set a validity period of 15 minutes.

[0085] S214. Determine preset dynamic desensitization rules for target sensitive data according to the target role authority.

[0086] The target role permissions are the permission configuration set required for a specific task node. The target sensitive data is the data object that needs access control and desensitization. The preset dynamic desensitization rules are data desensitization strategies dynamically generated according to the permission level. Desensitization rules define the processing methods of data fields, such as masking, replacement, encryption, etc. For example, for customer mobile phone numbers, different permission levels correspond to different desensitization rules: fully visible, hiding the last 4 digits, only showing the first 3 and last 4 digits, etc.

[0087] The system determines the desensitization strategy through the rule engine. First, load the desensitization rule template library, which includes the basic rule set and the business rule set. The basic rule set defines the desensitization methods for common data types, such as ID card number, mobile phone number, bank card number, etc. The business rule set defines the desensitization requirements for specific business scenarios. The system analyzes the three dimensions of the target role's permissions: data access level, operation permission type, and timeliness constraints, and generates a desensitization rule decision tree. According to the decision tree, match the appropriate desensitization rule combination, which includes: field desensitization method, desensitization parameter configuration, and exception processing conditions. Serialize the generated desensitization rules into configuration objects and store them in the rule cache.

[0088] S215. During the data access process to the target sensitive data, the target sensitive data is desensitized according to the preset dynamic desensitization rule.

[0089] The data access process refers to the process of the system reading and returning sensitive data. The preset dynamic desensitization rules are a set of predefined data processing strategies. Desensitization is the process of deforming or hiding sensitive data. The target sensitive data is the original data object that needs to be protected. For example, when accessing customer information, the system desensitizes the mobile phone number "13812345678" to "138****5678" and the ID number "440123199001011234" to "440123********1234" according to the rules.

[0090] The system implements real-time desensitization processing at the data access layer. First, a data access interceptor is built to intercept all query requests for sensitive data. The interceptor obtains the desensitization rule configuration for the current session from the context. The query result set is processed line by line, and the corresponding desensitization method is applied to each data field. Desensitization method implementations include: regular expression replacement, AES encryption, character masking, randomization processing, etc. The system uses a parallel stream processing framework to improve desensitization efficiency and shards the data set for parallel processing. After processing, the desensitized data is reassembled into a result set and returned. At the same time, the desensitization operation log is recorded, including desensitization rule ID, processing time, data volume and other information. The processing delay of the entire desensitization process is controlled within 50 milliseconds.

[0091] The following is a more detailed description of the process of the method provided by this implementation. Figure 3, which is another flow chart of the method for dynamic desensitization of sensitive data in a cloud environment in an embodiment of the present application.

[0092] S301. Obtain the historical audit log data of the target user within a preset time period, wherein the historical audit log data includes the target user's access time, access frequency, access location, data operation type, and task node conversion record in the task chain permission dependency graph to the target sensitive data.

[0093] The target user is a specific user account in the system that needs to be analyzed for behavior. The preset time period is a fixed time range for analysis, usually nearly 90 days. Historical audit log data is a record of all operations of users in the system. The access time records the specific timestamp of each data access. The access frequency counts the number of visits per unit time. The access location includes IP address and geographic location information. Data operation types include query, modification, deletion and other operation classifications. The task node transition record describes the node jump order of the user in the task process. For example, the audit log of a credit approval officer contains: {Timestamp: 2024-01-0110:00:00, IP: 192.168.1.100, Location: Headquarters Building, Operation: Query Customer Information, Task Node: Risk Assessment}.

[0094] The system obtains historical data through the audit log collection module. First, determine the data collection time range and set it to 90 days before the current time. Construct log query conditions: user ID, time range, and operation type list. Batch read qualified log records from the distributed log storage system. Clean and format the original log data: parse the JSON format log content, extract key fields, unify the time format to UTC timestamp, standardize the IP address format, and convert the geographic location code. Sort the log records in chronological order and calculate the access frequency indicators: number of visits per hour, number of visits per day, and distribution of visits on weekdays and non-workdays. Store the processed log data in the analysis database and create an index to support fast query.

[0095] S302: Train a machine learning model based on the historical audit log data to obtain a normal access behavior benchmark model of the target user.

[0096] Historical audit log data is a collection of time-series records of user behavior. Machine learning models are behavioral pattern recognizers obtained through data training. The normal access behavior baseline model defines the standard behavioral characteristics of users. The training process is a computational process of learning behavioral patterns from historical data. For example, the model learns that users usually access the system from a fixed office location between 9:00 and 18:00 on weekdays, process an average of 10 transactions per hour, and 90% of operations follow standard task processes.

[0097] The system adopts a multi-level behavior modeling approach. First, feature engineering is performed: extracting time features (hourly distribution, weekly distribution, holiday patterns), spatial features (IP address segments, geographic location sets), operation features (operation type distribution, data access patterns), and task features (node ​​conversion sequences, processing time distribution). Construct a behavior vector space, use One-Hot encoding to process category features, and use normalization to process numerical features. Use an ensemble learning method to train the model: use Isolation Forest to detect anomalies, use LSTM network to learn time series patterns, and use Gaussian mixture model (GMM) to establish behavior distribution. Model training uses a cross-validation method, using 80% data for training and 20% for validation. Save the trained model parameters as a benchmark model file.

[0098] Specifically, the model can compare and analyze the input target user's real-time access behavior data with the normal behavior pattern learned from historical audit log data, and output an evaluation result to indicate whether the current real-time access behavior conforms to the normal behavior pattern. This evaluation result can be in the form of a score, a category label (such as normal or abnormal), or a probability value (indicating the probability of normal behavior). For example, the model may output a score between 0 and 1. When this score is close to 1, it means that the real-time access behavior is very similar to the normal behavior pattern, and when it is close to 0, it means that the difference is large. When this score is lower than the preset threshold (such as 0.8), it is considered that the deviation between the real-time access behavior and the normal behavior exceeds the preset threshold, thereby triggering the subsequent adjustment of the desensitization level in the preset dynamic desensitization rule to the preset level. The input data of the normal access behavior benchmark model is the historical audit log data of the target user within a preset time period (usually nearly 90 days). These data record in detail the various access information of the target user to the target sensitive data, including access time (specific to timestamp), access frequency (number of visits per hour, day, etc.), access location (IP address and corresponding geographic location), data operation type (such as query, modification, deletion, etc.) and task node conversion records in the task chain permission dependency graph. The training data is the historical audit log data after cleaning and formatting. Specific cleaning and formatting operations include: parsing the log content in JSON format to extract key information, such as user ID, time range, operation type, etc.; converting the time format to UTC timestamp to facilitate subsequent time series analysis; standardizing IP addresses to make them in a unified format; converting geographic location information into a specific encoding format, etc. The input data is used to determine whether the target user conforms to normal behavior patterns, and an evaluation result can be output, such as a score or a category label (such as normal or abnormal).

[0099] S303. Evaluate the real-time access behavior of the target user according to the normal access behavior benchmark model. When the deviation between the real-time access behavior and the normal access behavior exceeds a preset threshold, adjust the desensitization level in the preset dynamic desensitization rule to a preset level.

[0100] The normal access behavior baseline model is a mathematical model that describes the standard behavior characteristics of users. Real-time access behavior is the characteristics of the operations that users are currently performing. The evaluation process is to calculate the degree of difference between the current behavior and the baseline model. The preset threshold is the standard value for judging abnormal behavior. The desensitization level is the intensity level of data protection, which is divided into multiple levels from low to high. The preset level is the protection level enabled after anomalies are found. For example, when a user accesses a large amount of sensitive data at unusual times, the system detects the anomaly and increases the desensitization intensity.

[0101] The system monitors and evaluates user behavior in real time. A real-time feature extractor is built to calculate the behavioral feature vector of the last 5 minutes every 60 seconds. The feature vector is input into the benchmark model to calculate the anomaly score: time dimension score (access probability at the current time point), space dimension score (location matching degree), operation dimension score (operation frequency deviation), and task dimension score (process standardization degree). The weighted average method is used to calculate the comprehensive anomaly score, with weights of 0.3, 0.2, 0.3, and 0.2 respectively. The anomaly threshold is set to 0.8, and the desensitization level adjustment is triggered when the comprehensive score exceeds the threshold. The desensitization level adjustment follows the preset rules: the anomaly score increases by one level in the [0.8-0.9] range, increases by two levels in the [0.9-1.0] range, and increases to the highest level when it exceeds 1.0. The adjusted desensitization rules take effect immediately, and the duration is set to 30 minutes.

[0102] S304: Collect the current access data of the target user in real time, and input the data into the normal data access behavior benchmark model for calculation to obtain an abnormality score of the current access behavior.

[0103] The target user is a specific user account that needs to be monitored in the system. The current access data is the system operation information that the user is currently performing, including real-time data such as access time, access address, and operation type. The normal data access behavior benchmark model is a user standard behavior model obtained through machine learning training. The abnormality score is a numerical indicator that measures the degree to which the current behavior deviates from the normal pattern, and the value range is [0, 1]. For example, a user downloads customer data in batches from an unknown IP address at 3 a.m., and the system calculates an abnormality score of 0.95, indicating that the behavior is highly abnormal.

[0104] The system obtains user behavior data through a real-time data collection pipeline. The collection module triggers data collection every 5 seconds and obtains the following information through the application program interface (API): user session information (session ID, login time, terminal type), access data (request URL, HTTP method, request parameters), and system information (CPU usage, memory usage, network traffic). Feature extraction is performed on the collected data: short-term features (operation frequency and data access volume in the last minute), medium-term features (operation type distribution and access mode in the last 10 minutes), and session features (accumulated number of operations and duration of the current session) are calculated. The extracted feature vector is input into the benchmark model to calculate the anomaly scores of four dimensions: time dimension (operation probability at the current time point), spatial dimension (credibility of access location), behavioral dimension (standardization of operation sequence), and resource dimension (rationality of resource use). The weighted summation method is used to calculate the final anomaly score: anomaly score = Σ (dimension weight × dimension score), with weights of 0.3, 0.2, 0.3, and 0.2 respectively.

[0105] S305: Adjust access rights to the target sensitive data according to the abnormality level score.

[0106] The abnormality score is a quantitative behavioral risk indicator. Target sensitive data is important business data that needs to be protected. Access rights are a set of rules that define the scope of user operations on data. Permission adjustment is the process of dynamically changing data access control policies based on risk levels. For example, when the abnormality score reaches 0.8, the system reduces the user's access rights to customer sensitive information from "full access" to "read-only access" and increases the intensity of field desensitization.

[0107] The system implements a risk-based adaptive access control mechanism. First, define the abnormality level classification: normal ([0-0.6]), mild abnormality ([0.6-0.8]), moderate abnormality ([0.8-0.9]), severe abnormality ([0.9-1.0]). For each abnormality level, set the corresponding permission adjustment strategy: enable operation auditing in case of mild abnormality and record detailed operation logs; reduce data access permissions in case of moderate abnormality, reduce write permissions to read-only permissions, and increase the field desensitization level; temporarily freeze sensitive operation permissions in case of severe abnormality, retain only basic query permissions, and use the highest level of desensitization for all sensitive fields. Permission adjustment execution process: obtain the user's current permission configuration from the permission configuration library, calculate new permission rules based on the abnormality level, update the permission configuration through the distributed cache, and set the permission effective time to 30 minutes. At the same time, start the permission recovery timer, and automatically restore the original permission settings after the user behavior returns to normal (the abnormality score is lower than 0.6 for 10 minutes). All permission adjustment operations are recorded in the audit log, including adjustment time, trigger reason, permission change details and other information.

[0108] It can be understood that steps S301 to S305 can be performed after step S215.

[0109] The following describes the dynamic desensitization system for sensitive data in a cloud environment in the embodiment of the present invention from the perspective of hardware processing. Figure 4 , which is a schematic diagram of the physical device structure of a dynamic desensitization system for sensitive data in a cloud environment in an embodiment of the present application.

[0110] It should be noted that Figure 4 The structure of the sensitive data dynamic desensitization system in the cloud environment shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0111] like Figure 4 As shown, the dynamic desensitization system for sensitive data in a cloud environment includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 402 or the program loaded from the storage part 408 to the random access memory (RAM) 403, such as executing the method described in the above embodiment. Various programs and data required for system operation are also stored in RAM 403. CPU 401, ROM 402 and RAM 403 are connected to each other through bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0112] The following components are connected to the I / O interface 405: an input section 406 including an audio input device, a button switch, etc.; an output section 407 including a liquid crystal display (LCD) and an audio output device, an indicator light, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed so that a computer program read therefrom is installed into the storage section 408 as needed.

[0113] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 409, and / or installed from a removable medium 411. When the computer program is executed by the central processing unit (CPU) 401, various functions defined in the present invention are performed.

[0114] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0115] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings.

[0116] Specifically, the dynamic desensitization system for sensitive data in a cloud environment of this embodiment includes a processor and a memory. A computer program is stored in the memory. When the computer program is executed by the processor, the dynamic desensitization method for sensitive data in a cloud environment provided in the above embodiment is implemented.

[0117] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the sensitive data dynamic desensitization system in the cloud environment described in the above embodiments; or may exist independently without being assembled into the sensitive data dynamic desensitization system in the cloud environment. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of the sensitive data dynamic desensitization system in the cloud environment, the sensitive data dynamic desensitization system in the cloud environment implements the sensitive data dynamic desensitization method in the cloud environment provided in the above embodiments.

[0118] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0119] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0120] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. A method for dynamic desensitization of sensitive data in a cloud environment, characterized in that: A sensitive data dynamic desensitization system applied in a cloud environment, the method comprising: Generate a task chain permission dependency graph based on the role permission combination, wherein the task chain permission dependency graph includes the permission inheritance and transfer relationship between each task node, and the role permission combination includes the viewing permission level, data operation permission and timeliness constraint; According to the current task node position of the target user, the next task operation of the target user is predicted in the task chain permission dependency graph, and the target role permission corresponding to the next task operation is preloaded; Determine preset dynamic desensitization rules for target sensitive data according to the target role permissions; During the data access process of the target sensitive data, the target sensitive data is desensitized according to the preset dynamic desensitization rules.

2. The method according to claim 1, characterized in that Before the step of generating a task chain permission dependency graph according to the role permission combination, the method further includes: Acquire the operation time series data of the target user, and extract the standard work task chain of the target user according to the operation time series data; Each task node of the standard work task chain is labeled with a role permission combination corresponding to the task node.

3. The method according to claim 2, characterized in that The step of obtaining the target user's operation time series data and extracting the target user's standard work task chain according to the operation time series data specifically includes: Performing time series analysis on the operation time series data to identify repeated operation patterns in the operation time series data; Determine an operation sequence in the repeated operation mode whose occurrence frequency exceeds a preset threshold as a candidate task chain; The task similarity is calculated for the candidate task chains, and the candidate task chains with similarity greater than a preset similarity threshold are merged to obtain the standard work task chain of the target user. The task similarity is determined by calculating the matching degree of the operation type, operation object and operation time of each task node in the candidate task chain.

4. The method according to claim 2, characterized in that: The step of generating a task chain permission dependency graph according to the role permission combination specifically includes: Obtain the role authority combination corresponding to each task node in the standard work task chain; Analyze the role authority combination of adjacent task nodes in the standard work task chain to generate an authority correlation matrix; Determine the permission inheritance relationship between task nodes according to the permission association matrix, and generate an initial permission dependency graph; Analyzing the permission transfer relationship between non-adjacent task nodes according to the initial permission dependency graph, wherein the permission transfer relationship indicates that permissions can be transferred through intermediate task nodes; The permission transfer relationship is updated to the initial permission dependency graph to form a task chain permission dependency graph.

5. The method according to claim 1, characterized in that The step of predicting the next task operation of the target user in the task chain authority dependency graph according to the current task node position of the target user specifically includes: Obtain the historical operation track of the target user in the task chain permission dependency graph; Analyze the association pattern between the task nodes in the historical operation trajectory to obtain the association relationship between the task nodes; Determine a set of predicted task operations that the target user may perform according to the association relationship and the task node position where the target user is currently located; Based on the historical execution frequency of each of the task nodes in the predicted task operation set, the task node with the highest execution probability is selected as the next task operation.

6. The method according to claim 1, characterized in that After the step of desensitizing the target sensitive data according to the preset dynamic desensitization rule during the data access process of the target sensitive data, the method further includes: Obtaining historical audit log data of the target user within a preset time period, the historical audit log data including the target user's access time, access frequency, access location, data operation type to the target sensitive data, and task node conversion records in the task chain permission dependency graph; Training a machine learning model based on the historical audit log data to obtain a normal access behavior benchmark model of the target user; The real-time access behavior of the target user is evaluated according to the normal access behavior benchmark model. When the deviation between the real-time access behavior and the normal access behavior exceeds a preset threshold, the desensitization level in the preset dynamic desensitization rule is adjusted to a preset level.

7. The method according to claim 6, characterized in that After the step of training the machine learning model based on the historical audit log data to obtain a normal data access behavior benchmark model of the target user, the method further includes: Collecting the current access data of the target user in real time, and inputting it into the normal data access behavior benchmark model for calculation to obtain the abnormality score of the current access behavior; The access permission to the target sensitive data is adjusted according to the abnormality degree score.

8. A dynamic desensitization system for sensitive data in a cloud environment, characterized in that: The sensitive data dynamic desensitization system in the cloud environment includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the sensitive data dynamic desensitization system in the cloud environment to execute the method described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on a dynamic desensitizing system for sensitive data in a cloud environment, the dynamic desensitizing system for sensitive data in a cloud environment executes a method as described in any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product runs on a dynamic desensitization system for sensitive data in a cloud environment, the dynamic desensitization system for sensitive data in a cloud environment executes the method as described in any one of claims 1 to 7.

Images